AI in private credit: Strengthening the reconciliation function

The more manual a process is, the more inherent risk there is. Nowhere is this truer than in private credit, where high volumes of non-standard data complicate the collection, sharing and validation of loan data. As one of the newer markets in the loan industry, with less maturity of data standardization and processes across market participants, private credit is particularly susceptible to data inaccuracy and manual errors.

The combination of unstructured data, a lack of standardization, and the opacity and illiquidity of the assets themselves has resulted in a fragile, resource-intensive workflow that isn't built to scale. But scale it must, as the asset class faces strong projected growth in the coming years. 

At S&P Global Market Intelligence, we are applying AI’s unique capabilities to the reconciliation workflow to transform the efficiency, accuracy and turnaround for this critical function.

Achieving reconciliation at scale

Our managed service for reconciliation is built on WSO Software, a platform that delivers portfolio insights, workflow automation, and deal administration for diversified credit portfolios, CLOs, syndicated and leveraged loans. Over the past year, WSO’s AI tools have enabled our services teams to manage the reconciliation of cash payments and position-level data faster and more accurately than ever before.

Prior to launching AI functionality, our team was able to automate 50% of reconciliation activities across a global customer base. Today, we are automating more than 90% of those cash transactions. That's on an annual basis across over 4.5 million cash payments globally, demonstrating that this technology is capable of automating reconciliation processes at virtually any scale.

Super-human analytic capabilities

Matching cash is often a needle-in-a-haystack exercise with many-to-one and many-to-many combinations that humans are slower decipher. Automating the function is complicated further by complex requirements for matching issuer names and security IDs and non-standard data formats for external cash files.

AI can manage these complex equations and find the multi-variable matches. It can even identify cash that is posted to the wrong portfolio account and determine how to resolve the break.  

With an AI assist, the S&P Global Market Intelligence team can reconcile over 8,000 active ledgers representing $1.6 trillion AUM daily and recover 100% of cash payments at quarter-end. But it's the speed with which this high-volume task can now be performed that our clients appreciate most: the process is now completed within a three-day period compared to the team's previous best efforts of ten business days.

Smarter risk mitigation

Automation has enabled us to focus on the true exceptions—the large, material events that represent the greatest risk for our customers, such as missing or incorrect payments. With the vast majority of the reconciliation process fully automated, our team can focus on the small proportion of exceptions and bring more care and diligence to reconciling those discrepancies.

We are also using AI to scrape and cleanse the required data points, including issuer names, security IDs, date formats, currency formats, etc., from disparate file formats that contain position and cash transactions. By leveraging LLMs for querying and validating data against loan reference data tables, we are enabling our team to enhance data accuracy and compliance for the clients who rely on their services.

Paired with an API that makes reconciled matches available to downstream processes, this innovation significantly improves and accelerates the reconciliation process. And this is just the start in terms of streamlining processes, enhancing the user experience and making it easier to focus on the true risk in a portfolio. 

The future of AI in private credit

At S&P Global Market Intelligence, we moved early and invested deeply in AI, and we are already seeing its impact on reconciliation, which has long been one of the most challenging aspects of the private loan lifecycle. But this is only the beginning. Soon our clients will benefit from self-learning AI that continually analyzes the way human operatives resolve exceptions, (such as multiple naming conventions for a single loan), so that it can apply advanced reasoning to these issues without human intervention. We are also working on AI capable of reconciling cash payments even when they do not tie back to assets (loans, bonds and equities) tracked in the system.

Ultimately, AI is taking us to a future where lenders can consume and standardize data in any format and from any point of origin. For private credit, the impact of this reality can't be overstated.

Read the other installments in our AI in lending blog series:

• AI in Private Credit: Raising the Bar for STP
• AI in Private Credit: Surfacing Deeper Data Insights
• AI in Private Credit: Reshaping the User Experience